Reservoir porosity is a critical characteristic for determining reservoir performance. At the moment, the correlation with porosity of core analysis is typically chosen. The increased logging data and porosity create a multiple linear regression model to estimate reservoir porosity. Ignoring logging data with poor correlation may result in some information leakage of formation porosity, and multicollinearity between variables may lead to regression model instability when employing logging data for multivariate comprehensive analysis, increasing prediction error. In light of the aforementioned issues, logging data indicating the formation's acoustic, electrical, and radioactive properties are carefully chosen, and the porosity of the Yan 9 part of the research area is forecasted using a BP neural network and linear regression. The results find that the accuracy of the BP neural network's porosity prediction results is higher, and it is clearly superior than the results of regression analysis prediction. This method has a good effect on quantitative prediction of reservoir porosity.